Diversity-aware Channel Pruning for StyleGAN Compression
This addresses the challenge of making StyleGAN more practical for applications by improving compressed model diversity, though it is incremental as it builds on prior compression methods.
The paper tackles the problem of compressing StyleGAN to reduce computational cost while preserving sample diversity, proposing a channel pruning method that enhances diversity and achieves state-of-the-art FID scores with half the training iterations.
StyleGAN has shown remarkable performance in unconditional image generation. However, its high computational cost poses a significant challenge for practical applications. Although recent efforts have been made to compress StyleGAN while preserving its performance, existing compressed models still lag behind the original model, particularly in terms of sample diversity. To overcome this, we propose a novel channel pruning method that leverages varying sensitivities of channels to latent vectors, which is a key factor in sample diversity. Specifically, by assessing channel importance based on their sensitivities to latent vector perturbations, our method enhances the diversity of samples in the compressed model. Since our method solely focuses on the channel pruning stage, it has complementary benefits with prior training schemes without additional training cost. Extensive experiments demonstrate that our method significantly enhances sample diversity across various datasets. Moreover, in terms of FID scores, our method not only surpasses state-of-the-art by a large margin but also achieves comparable scores with only half training iterations.